Reducing Uncertainty in Space-Based Observations through Coordinated Collection
The Lumos Collection Recommender takes advantage of all available space sensors to characterize critical deep space threat. It significantly reduces the uncertainty of a resident space object’s (RSO) state by creating a set of coordinated collections across multiple space sensors. The Collection Recommender leverages both a Markov chain Monte Carlo (MCMC) and a Genetic Algorithm to provide an operator with multiple solutions that meet mission timelines and asset availability. Users can specify a targeted maximum RSO uncertainty (covariance) at a future time. This ensures the RSO’s state is well-defined for a future course of action.
The Lumos Collection Recommender Offers:
- Customizable Interactive Environment for Collection Optimization: Our customized OpenAI Gym environment models both spacecraft dynamics and target uncertainties using an Unscented Kalman Filter to determine optimal collection windows for multiple sensors to view the same target.
- Sensor Agnostic Near-Real-Time Processing: Lumos is sensor agnostic and currently performs collection optimization across all Space Surveillance Network (SSN) space-based sensors. Sensor specific capabilities are captured in configuration files and adding new constellations only requires configuration updates, and no code updates.
- Autoscaling Microservices: Lumos is a microservice architecture that can autoscale with dynamic cloud resources to match demand. This allows the Collection Recommender to quickly compute tasking during critical world events.
- Small Business Innovation and Cost: Lumos was awarded as a part of Space Pitch Day 2019 by General Thomson and Dr. Roper. The Collection Recommender was developed for the standard award value of $750K.